A study in JACC yesterday looked at the inclusion of three new attributes to predict in-hospital mortality following PCI that has implications for public reporting of hospital performance. In Massachusetts, where the study was conducted, reporting of in-hospital PCI mortality rates has been required since 2003 through the NCDR CathPCI Registry. However, since the CathPCI Registry was not built for public reporting and not intended to identify high-risk clinical scenarios, physicians in 2006 recommended inclusion of three additional attributes, which they deemed “compassionate use” (CU) measures. These measures are: coma on presentation, active hemodynamic support during PCI and cardiopulmonary resuscitation at PCI initiation. The purpose of the study was to see if including these measures was feasible and would improve the prediction model for in-hospital PCI mortality.

Researchers divided patients into two categories: those presenting with STEMI or cardiogenic shock (which they called the SOS group), and all others (the non-SOS group). The findings were definitive:

• The unadjusted in-hospital mortality rate was 15.6 times higher for CU patients vs. non-CU patients in the SOS group;
• CU patients only represented 1.6% of all SOS patients, but represented 21% of the overall mortality after PCI for the SOS group; and
• Being designated CU was associated with an odds ratio for in-hospital death of 27.3 relative to the non-CU SOS patients, after adjusting for other known predictors of in-hospital mortality.

The authors conclude:

“The Massachusetts experience demonstrates that a small proportion of patients at extremely high risk of in-hospital mortality can be identified using objective, pre-procedure clinical factors that had not been previously collected as part of traditional quality monitoring efforts. Incorporation of these CU covariates in risk-adjustment models led to significant improvements in model performance as well as reclassification of predicted risk in a substantial proportion of cases.”

Editorial
Eric Peterson, MD, MPH, writes the editorial that accompanies the article. He’s right on in his comments. With patients and the government increasingly demanding transparency in health care outcomes, there is a true risk of unintended consequences if we do not do due diligence in taking into account as many predictors of adverse outcomes as necessary. There are plenty of anecdotes of physicians refusing to perform CABG or PCI because they simply don’t want to have a death on their outcomes report by caring for patients whose clinical status is so extreme that their chance of surviving the hospitalization after the emergent CABG or PCI might be less than 30%.

Although the NCDR has a robust risk adjustment model to “level the playing field” to take account of these very ill patients, Resnic, et al., offer CU risk adjustment measures that appear to be an improvement in accurately risk-adjusting the severity of illness for these infrequent but critically ill patients. The use of the CU measures might hope to mitigate against the negative unintended consequences of public reporting – that is to say, physicians would be more willing to take these ill patients to the cath lab for emergent PCI rather than refusing to do so.

The challenge of the CU risk-adjusted measures is for its accurate “coding” by the data analysts when submitting their registry reports. A robust auditing system of all patients deemed to meet CU criteria is necessary to assure accuracy in the coding of these patients.

What do you think of the study findings? Should (and if so, how) CU criteria be incorporated into public reporting efforts?